First thoughts on the incorporation of cultural variables into predictive modelling

(1)

First thoughts on the incorporation of cultural variables into predictive

modelling

Verhagen, P.; Kamermans, H.; Leusen, M. van; Deeben, J.; Hallewas, D.P.; Zoetbrood, P.; Verhagen Ph

Citation

Verhagen, P., Kamermans, H., Leusen, M. van, Deeben, J., Hallewas, D. P., & Zoetbrood, P. (2007).

First thoughts on the incorporation of cultural variables into predictive modelling. In Case Studies in Archaeological Predictive Modelling (pp. 203-210). Amsterdam: Leiden University Press.

Retrieved from https://hdl.handle.net/1887/17626

Version: Not Applicable (or Unknown)

License: Leiden University Non-exclusive license Downloaded from: https://hdl.handle.net/1887/17626

Note: To cite this publication please use the final published version (if applicable).

(2)

CASE STUDIES IN ARCHAEOLOGICAL PREDICTIVE MODELLING

(3)

Archaeological Studies Leiden University

is published by Leiden University Press, the Netherlands Series editors: C.C. Bakels and H. Kamermans

Cover illustration: Philip Verhagen Cover design: Medy Oberendorff Lay out: Philip Verhagen ISBN 978 90 8728 007 9 NUR 682

or transmitted, in any form or by any means (electronic, mechanical, photocopying, recording or otherwise) without the written permission of both the copyright owner and the author of the book.

(4)

TABLE OF CONTENTS

PREFACE... ... 9

CHAPTER 1 A Condensed History of Predictive Modelling in Archaeology ... 13

1.1. INTRODUCTION ... 13

1.2. THE ORIGINS OF ARCHAEOLOGICAL PREDICTIVE MODELLING ... 14

1.3. GIS IN ARCHAEOLOGY... 15

1.4. THE CONTROVERSY ON PREDICTIVE MODELLING... 17

1.5. PREDICTIVE MODELLING IN CULTURAL RESOURCE MANAGEMENT... 17

1.6. PREDICTIVE MODELLING IN THE NETHERLANDS... 18

1.7. THE BBO PREDICTIVE MODELLING PROJECT... 20

PART 1: PRACTICAL APPLICATIONS... 27

CHAPTER 2 The Use of Predictive Modeling for Guiding the Archaeological Survey of Roman Pottery Kilns in the Argonne Region (Northeastern France) ... 29

2.2. ARCHAEOLOGICAL CONTEXT ... 30

2.3. AREA DESCRIPTION... 32

2.4. THE FIRST PREDICTIVE MODEL ... 34

2.5. THE SECOND PREDICTIVE MODEL ... 35

2.6. THE FINAL MODEL... 36

2.7. CONCLUSIONS... 38

CHAPTER 3 The hidden reserve. Predictive modelling of buried archaeological sites in the Tricastin- Valdaine region (Middle Rhône Valley, France) ... 41

3.2. THE PREDICTIVE MODEL ... 42

3.3. THE PREDICTIVE MODEL: METHODS APPLIED... 47

3.4. THE PREDICTIVE MODEL: RESULTS OF SITE LOCATION ANALYSIS... 50

3.5. EXTRAPOLATING SITE DENSITIES... 62

CHAPTER 4 Quantifying the Qualified: the Use of Multicriteria Methods and Bayesian Statistics for the Development of Archaeological Predictive Models... 71

4.2. MULTICRITERIA DECISION MAKING AND ITS RELEVANCE TO PREDICTIVE MODELING ... 72

4.3. BAYESIAN STATISTICS AND PREDICTIVE MAPPING... 77

4.4. APPLICATION: THE PREDICTIVE MAP OF EDE... 81

PART 2: ARCHAEOLOGICAL PROSPECTION, SAMPLING AND PREDICTIVE MODELLING ... 93

(5)

TABLE OF CONTENTS

CHAPTER 5 Establishing optimal core sampling strategies: theory, simulation and practical

implications... 95

5.2. CORE SAMPLING: THE BASICS... 95

5.3. STATISTICAL BACKGROUND ... 96

5.4. ESTABLISHING AN OPTIMAL CORE SAMPLING STRATEGY: THE CASE OF ZUTPHEN-OOIJERHOEK ... 97

CHAPTER 6 Prospection strategies and archaeological predictive modelling... 101

6.2. PROSPECTION STRATEGIES... 101

6.3. CONTROLLING SURVEY BIASES ... 103

6.4. INTERSECTION PROBABILITY... 104

6.5. SURVEY INTENSITY AND TESTING OF PREDICTIVE MODELS... 106

6.6. DETECTION PROBABILITY... 107

6.7. LARGE OR SMALL INTERVENTIONS?... 108

CHAPTER 7 Predictive models put to the test ... 115

7.1.1 BACKGROUND ... 115

7.1.2 A NOTE ON TERMINOLOGY... 115

7.1.3 EXPERT JUDGEMENT TESTING: AN EXAMPLE FROM PRACTICE... 116

7.2. MODEL PERFORMANCE ASSESSMENT... 119

7.2.1 GAIN AND RELATED MEASURES ... 120

7.2.2 MEASURES OF CLASSIFICATION ERROR ... 121

7.2.3 PERFORMANCE OPTIMISATION METHODS ... 125

7.2.4 PERFORMANCE ASSESSMENT OF DUTCH PREDICTIVE MODELS ... 126

7.2.5 COMPARING CLASSIFICATIONS... 128

7.2.6 COMPARING CLASSIFICATIONS: AN EXAMPLE FROM PRACTICE... 129

7.2.7 SPATIAL AUTOCORRELATION AND SPATIAL ASSOCIATION ... 132

7.2.8 SUMMARY AND DISCUSSION... 133

7.3. VALIDATION OF MODEL PERFORMANCE ... 136

7.3.1 SIMPLE VALIDATION TECHNIQUES ... 137

7.3.2 SIMPLE VALIDATION AND PREDICTIVE MODELLING... 139

7.4. STATISTICAL TESTING AND PREDICTIVE MODELS... 141

7.4.1 WHY USE STATISTICAL TESTS? ... 141

7.4.2 HOW TO TEST RELATIVE QUALIFICATIONS ... 143

7.5. COLLECTING DATA FOR INDEPENDENT TESTING... 145

7.5.1 PROBABILISTIC SAMPLING ... 146

7.5.2 SURVEY BIAS AND HOW TO CONTROL FOR IT... 148 7.5.3 USING THE ARCHIS DATABASE FOR PREDICTIVE MODEL TESTING . 149

(6)

TABLE OF CONTENTS

7.5.4 TESTING THE ENVIRONMENTAL DATA ... 152

7.5.5 CONCLUSIONS ... 153

7.6. THE TEST GROUND REVISITED... 153

7.6.1 MODEL TYPES AND APPROPRIATE TESTING METHODS... 153

7.6.2 TOWARDS AN ALTERNATIVE FORM OF PREDICTIVE MAPPING: RISK ASSESSMENT AND THE USE OF AREA ESTIMATES ... 156

7.7. CONCLUSIONS AND RECOMMENDATIONS... 159

7.7.1 CONCLUSIONS ... 159

7.7.2 RECOMMENDATIONS... 162

PART 3: ALTERNATIVE WAYS OF PREDICTIVE MODELLING ... 169

CHAPTER 8 Modelling Prehistoric Land Use Distribution in the Río Aguas Valley (S.E. Spain) .... 171

8.2. ENVIRONMENTAL CONTEXT ... 174

8.3. ARCHAEOLOGICAL CONTEXT ... 174

8.4. AGRICULTURAL POTENTIAL OF THE RÍO AGUAS VALLEY ... 175

8.5. LAND SUITABILITY: A FUNCTION OF POTENTIAL AND ACCESSIBILITY.... 177

8.6. ESTIMATION OF LAND SURFACE NEEDED FOR AGRICULTURE ... 178

8.7. FINDING THE LAND ... 179

8.8. RESULTS ... 180

CHAPTER 9 Some considerations on the use of archaeological land evaluation ... 193

9.2. ENVIRONMENTAL CHANGE AND ITS CONSEQUENCES FOR LAND SUITABILITY... 194

9.3. TECHNOLOGICAL DEVELOPMENT: HYDRAULIC INFRASTRUCTURE... 196

9.4. THE HUMAN PERCEPTION OF SUITABILITY... 198

CHAPTER 10 First thoughts on the incorporation of cultural variables into predictive modelling ... 203

10.2. PREDICTIVE MODELLING AND ENVIRONMENTAL DETERMINISM... 204

10.3. CULTURAL VARIABLES: WHAT ARE THEY?... 205

10.4. HOW TO PROCEED?... 206

EPILOGUE WHITHER ARCHAEOLOGICAL PREDICTIVE MODELLING?... 211

SAMENVATTING... 215

(7)

203

CHAPTER 10 First thoughts on the incorporation of cultural

variables into predictive modelling

¹

Philip Verhagen, Hans Kamermans², Martijn van Leusen³, Jos Deeben⁴, Daan Hallewas⁴ and Paul Zoetbrood⁴INTRODUCTION

Predictive modelling is a technique used to predict archaeological site locations on the basis of observed patterns and/or assumptions about human behaviour (Kohler and Parker, 1986; Kvamme 1988;

1990). It was initially developed in the USA in the late 1970s and early 1980s where it evolved from governmental land management projects and is still regularly applied in cultural resources management. In the Netherlands, predictive modelling plays an important role in the decision making process for planning schemes on a municipal, provincial and national level.

However, in many other countries predictive modelling is far from being an accepted tool for archaeological heritage management (AHM), and even where it is used regularly, criticism is not uncommon (see e.g. Ebert, 2000; Whitley, in press; van Leusen et al., 2002). Much of this criticism is related to the uncritical application of so-called 'inductive' modelling techniques, in which the archaeological data set is used to obtain statistical correlations between the location of archaeological sites and environmental variables such as soil type, slope or distance to water. The performance of these models is in general not very good, partly because of the use of inappropriate statistical techniques, but mainly because of the biased nature of many archaeological data sets and the emphasis on environmental factors, which are easier to model than the more intangible social and cultural factors.

Wheatley (2003) even states that, as predictive modelling doesn't work very well, it shouldn't be used at all: "Archaeology should really face up to the possibility that useful, correlative predictive modelling will never work because archaeological landscapes are too complex or, to put it another way, too interesting". His argument is mainly directed against the use of biased archaeological data sets, that will lead to the development of biased models that will in turn inevitably produce a positive feedback loop of even more biased data sets, as it is common practice to spend funds for AHM on the areas of 'high archaeological value'.

These areas will become better and better known, whereas the areas that are designated a 'low value' on the predictive map will largely be ignored in (commercial) archaeological research.

Verhagen (in press) however shows that the creation of biased data sets is not just a problem of predictive modelling, but a more general characteristic of the way in which archaeological data is collected.

Most of the archaeological prospection done is not taking into account statistical sampling theory, and it can be suspected that many survey projects do not even have a strong archaeological hypothesis in mind. We believe that predictive modelling can serve as a means to make explicit the assumptions that are often made concerning the location preferences of prehistoric people. A predictive model should be based on a theory of site location preferences, that can be quantified and tested against (unbiased) archaeological data sets (see also

1 This paper was presented by Hans Kamermans at the CAA 2004 conference, held from 13-17 April 2004 in Prato, Italy, and will be published in its proceedings. The text of this paper was largely prepared by me in co-operation with Hans Kamermans and Martijn van Leusen, but as it is part of the research done for the BBO-project ‘Predictive Modelling’, the other participants in the project are given credit as co-authors.

2 Faculty of Archaeology, Leiden University 3 Institute of Archaeology, Groningen University

4 Rijksdienst voor het Oudheidkundig Bodemonderzoek, Amersfoort

(8)

204 CHAPTER 10

Whitley, in press). It is clear that the cultural component of these theories is at the moment virtually absent in predictive modelling practice. This paper intends to show that it is not impossible to include these variables into predictive modelling, and this will hopefully lead to further research into this subject.

10.2. PREDICTIVE MODELLING AND ENVIRONMENTAL DETERMINISM

The practice of predictive modelling for AHM is, at the moment, environmental deterministic in outlook and design. The predominant use of environmental input variables as archaeological site predictors, such as soil type, groundwater table, distance to open water and the like, has however been criticized on a number of occasions in academic literature (e.g. Wheatley 1993; 1996a; 2003; Gaffney and van Leusen, 1995).

The problems associated with environmentally based predictive modelling (van Leusen et al., 2002) can be summarized as follows:

- archaeological theorists reject an understanding of past human behaviour in purely ecological/economical terms, and argue that social and cognitive factors determine this behaviour to a large extent, and should therefore be additional predictors for the presence and nature of archaeological remains;

- the maximum gain (a measurement of the degree of effectiveness of the predictive archaeological model over a ‘by chance’ model) of current predictive models seems to be about 70% (Ebert, 2000;

Wheatley, 2003), which implies that a significant proportion of archaeological site locations cannot be predicted using purely environmental datasets; therefore, models based on environmental factors alone cannot be adequate tools for the prediction of archaeological site location.

- unfortunately, social and cognitive factors seem to be difficult to model, and have so far only be studied for a very limited range of questions, based on very specialised data sets (mostly relating to the ritual prehistoric landscapes of Wessex in England; e.g. Wheatley 1995; 1996b).

The American archaeologist Timothy Kohler observed this as early as 1988. “Why are the social, political, and even cognitive/religious factors that virtually all archaeologists recognize as factors affecting site location and function usually ignored in predictive modelling?” (Kohler, 1988:19). He gives the answer a few pages later: “Given the subtleties and especially the fluidity of the socio-political environment, is it any wonder that archaeologists have chosen to concentrate on those relatively stable, “distorting” factors of the natural environment for locational prediction?” (Kohler, 1988:21).

In essence, the situation has not changed since Kohler made these remarks. The present practice of predictive modelling is still very much environmentally deterministic. Cultural variables are not included in the models, resulting in predictions ultimately based on physical properties of the current landscape.

Practitioners of 'traditional' predictive modelling have mostly resisted the conclusion that their models will not be adequate because they lack the input of non-environmental data (e.g. Kvamme, 1997). It is not because they do not want to include non-environmental factors; the problem is that these variables are regarded as being too abstract and intangible for use in a predictive model. Such models, so the argument goes, will therefore not become any better by investing valuable research time in mapping cultural variables. Several publications have focused on this apparent impossibility to incorporate non-environmental variables in predictive modelling (Wheatley, 1996a; Stančič and Kvamme, 1999; and Lock 2000). As a consequence, very few studies are available where an attempt is made to improve the gain of a model by incorporating non- environmental factors. As a consequence, the effect of including cultural variables into predictive models can

(9)

205 CHAPTER 10

at the moment not be assessed. The current situation is therefore characterized by a fundamental criticism of the environmental deterministic approach, coupled to a very poor insight into the potential of using cultural variables in predictive modelling.

Ultimately, the theoretical basis needed for the development of culturally based predictive models seems to be underdeveloped. It is evident that many models of prehistoric land use have been proposed for local case studies, but they are usually not generalized for application in a predictive modelling context, and often have never been tested in a rigorous way. A typical example of this is found in the theories regarding the location of Linear Band Ceramic settlements, in which a strong cultural component is supposed to be present (see Gaffney and van Leusen, 1995), yet no predictive model based on this assumption has ever been made.

In conclusion, it may be suspected that the lack of progress in incorporating cultural variables into predictive modelling has less to do with the variables themselves, than with the geographic and interpretative models needed to operationalize them for predictive modelling. Many applications that claim to be exponents of cognitive archaeology, often framed in post-processual rhetoric, rely on the same techniques that are used for old-fashioned, processual studies, up to the extent where they might even be called ‘cognitive deterministic’.

10.3. CULTURAL VARIABLES: WHAT ARE THEY?

It is important to realize that, when we are speaking of cultural variables, we can think of two ways of obtaining them. The first one is to consider them as measurable attributes of the archaeological sample that are not related to an environmental factor. So, instead of measuring for each individual site its soil type, elevation, distance from water and so on, we need to ask which properties of the site itself can be measured. These include properties like site location, size, functional type and period of occupation. These variables are clearly the expression of forms of social behaviour, although the interpretation of the specific behaviour involved may be subject to discussion. For ease of reference, these variables will be denominated cultural variables sensu stricto. In themselves, these variables are not extremely difficult to obtain, but the problems of analysing and interpreting archaeological site databases are manifold and must be addressed before these properties can actually be used for predictive modelling.

The second approach to defining cultural variables is to identify features of the landscape itself that can be interpreted as having cultural significance, such as sacred springs. These can be referred as to as cultural landscape variables, and are not necessarily excluded from ‘traditional’ predictive modelling, but often are not recognized as constituting a ‘cultural’ variable. It can, in fact, be argued that all environmental variables have a cultural component, even though the emphasis in traditional predictive modelling is usually on subsistence economy rather than symbolic meanings.

In order to make further use of cultural variables in predictive modelling, it is necessary to transform these variables into continuous variables: for each single variable a value should be available at any location within the study area. This is generally not a problem when using environmental data sets like soil maps or digital elevation models. Archaeological sites however are mostly represented as points, or in some cases as areas of a very limited extent. Similarly, landscape features that are considered to have cultural significance are in practice often also regarded as point-like, or at best linear in nature. A transformation is therefore necessary to use point-like or linear objects for predictive modelling. Two types of GIS techniques are currently available to perform this transformation: distance zonation and line-of sight analysis.

(10)

206 CHAPTER 10

Distance zonation is customarily performed in environmental predictive modelling to obtain continuous variables from environmental features that are either linear (like rivers or coastlines) or point-like (springs).

In some cases, cost surfaces (also known as friction surfaces or effort models) are calculated by assigning a weight to landscape features according to their supposed accessibility. This technique is applicable to environmental as well as cultural variables.

Distance decay models are used less often, and are based on demographic and/or political-economic models borrowed from human geography (e.g. Renfrew and Level, 1979). These models are specifically relevant for cultural variables sensu stricto, as they make it possible to incorporate the notion of interdependence of settlements (see e.g. Favory et al., 2003).

A number of studies have appeared in recent years using line-of-sight analysis as a technique for obtaining continuous cultural variables, amongst others in attempts to demonstrate the ritual and symbolic meaning of the placement of monuments such as long barrows (Wheatley, 1995; Gaffney et al., 1995).

However, this type of analysis is certainly not restricted to cultural variables.

A good example of the use of cultural variables sensu stricto and distance zonation is provided by Ridges (in press), who attempted to include the distance to rock art sites in a predictive model in NW Queensland (Australia) - and actually succeeded in improving the gain of the model. This success is probably due to the fact that the ritual sites used are fixed in space, and can be mapped with relative ease in the specific environmental situation. The rock art sites are typical examples of what Whitley (2000) refers to as ‘fixed point attractors’. The precise moment of their creation may be unknown, but their position and symbolic meaning remain stable during a long period of time, making them long-term attractors for human activity⁵.

In many other situations however, potential cultural variables are less stable, and cannot be mapped with ease. Examples of these include road networks, field systems, and the archaeological sites themselves, which all can have highly varying life-spans and may change in importance as attractors over time. In order to model the effects of long term land use development, it is necessary to use a technique that can deal with spatio-temporal variables, like dynamical systems modelling.

10.4. HOW TO PROCEED?

In order to remedy the current situation the following issues should be addressed:

- the identification of cultural variables that are significant for archaeological site location;

- the analysis of the utility of these variables for predictive modelling;

- the development and application of existing and new relevant modelling techniques; and

- the analysis of the performance of predictive models based on cultural variables compared to environmentally based models.

Following the recommendations in van Leusen et al. (2002), we suggest that four promising areas of research should be explored in order to improve on the current use of cultural variables in predictive modelling. These are:

5 in the case of Aboriginal rock art sites, it might even be a combination of ecological and cultural factors, as the sites are supposed to have been used as ‘markers’, indicating the presence of natural resources

(11)

207 CHAPTER 10

A systematic analysis of the archaeological records and their aggregation into culturally meaningful entities It is necessary to analyse what information can be extracted from existing archaeological databases that can be used in the definition of cultural variables. The aggregation of the archaeological contents of find spots into meaningful archaeological entities is currently not standardized. A possible solution could be to design an expert system that can be used for the classification of find spots. Apart from defining meaningful archaeological entities, the aggregation of multiple find spots into single archaeological sites is an important issue where the utility of the archaeological database for predictive modelling is concerned. Thirdly, a tendency can be observed recently to combine multiple archaeological sites into ensembles, which effectively constitutes a step away from the site level and towards a regional, landscape-based concept of archaeological entities (see also Kuna, 2000).

The main question here is: what types of aggregates can we distinguish, and can these be used as cultural variables sensu stricto?

Analysis of the logistic position of settlements

It is anticipated that one of the most important cultural variables that can be used is the logistic position of the archaeological site itself. It has been shown by many researchers that the position of a settlement in a logistic network determines to a large degree its size and duration of occupation (e.g. Durand- Dastès et al., 1998). The development of techniques to analyse the logistic position of settlements can be addressed by looking at recent work in human geography.

The continuity of the cultural landscape

The cultural landscape has a historical dimension that strongly influences its use and usability. The existing cultural landscape influences the positioning of new sites. Kuna (1998), for example, mentions the importance of remnants of past landscapes on settlement location choice. Bell et al. (2002) demonstrated how later settlement in their Central Italian study area avoids areas settled in an earlier phase but conforms to paths from that earlier phase. Techniques to perform the long-term diachronical analysis needed for this type of modelling have been developed (e.g. by the Archaeomedes project; van der Leeuw, 1998; Favory et al., 2003).

Line-of-sight analysis

In hilly areas and with certain site types that have a strong visual component (like burial mounds or megalithic tombs) line-of-sight analysis may be a type of analysis suitable for predictive modelling (see van Leusen, 2002: chapters 6 and 16). The techniques for performing this type of analysis are well established.

It will be noticed that the four research topics mentioned here all focus on cultural variables sensu stricto. A thorough investigation of the use of cultural landscape variables would primarily involve the development of a decision rule framework that will incorporate the perception of the landscape into predictive modelling. In itself, this is an issue that merits attention, but the establishment of decision rules has always been at the heart of predictive modelling and is covered by a wide range of studies already. It would however be useful to start thinking about ways to model the perception of the landscape, as has been done by Whitley

(12)

208 CHAPTER 10

(2000), who tried to model the attractivity of the landscape for specific (economic) activities of Native American hunter-gatherers (see also Whitley, in press).

10.5. CONCLUSIONS

In a recent article on the use and abuse of statistical methods in archaeological site location modelling Woodman and Woodward (2002) come to the following conclusion: “There has been much criticism of locational studies since they are often based largely on environmental criteria. However, before researchers attempt to incorporate the more intangible social, cognitive, political and aesthetic factors, it would be wise to employ the appropriate statistical techniques required to deal with the complexities which already exist in even the most basic tangible and quantifiable environmental criteria”.

Although we do not deny that many statistical problems still exist with regard to predictive modelling, we see no apparent reason why they should receive prime importance in further developing predictive modelling. In fact, the three main issues of statistical methodology, the development of adequate archaeological (and non-archaeological) data sets and the incorporation of non-environmental factors into the models are closely connected, and cannot be tackled in isolation. The papers presented in van Leusen and Kamermans (in press) show that new approaches to predictive modelling are starting to emerge, like exploring the potential of Bayesian statistical methods, using high resolution data for predictive modelling, and looking for ways to better embed predictive models into archaeological heritage management practice, for example by developing risk assessment methods. There is no doubt still a lot to do, and in this respect we have to disagree with Wheatley (2003) who argues that too much money is going into predictive modelling studies. He may be right that funding for GIS-related archaeological projects is mainly going into predictive modelling, but compared to the amount of money spent on all forms of prospection and excavation, investments made in predictive modelling seem relatively modest. Apart from that, investments for a thorough, scientific analysis of predictive modelling have been few and discontinuous.

We hope to have demonstrated that incorporating cultural variables into predictive modelling can be done, even though it is impossible to present a comprehensive overview in these few pages. It is up to the scientific community and public institutions to decide if this line of research is worth investing in. However, if the three issues mentioned above (statistical improvements, quality of the archaeological data set and the development of non-environmentally based models) are not tackled in the years to come, predictive modelling will remain to be criticized as a tool that is of dubious scientific quality, and not even capable of providing clear answers on where to spend money for archaeological research.

REFERENCES

Bell, T., A. Wilson and A. Wickham, 2002. ‘Tracking the Samnites: landscape and communications routes in the Sangro Valley, Italy’. American Journal of Archaeology 106 (2), pp. 169-186.

Durand-Dastès, F., F. Favory, J.-L. Fiches, H. Mathian, D. Pumain, C. Raynaud, L. Sanders and S. van der Leeuw, 1998. Des oppida aux métropoles. Archéologues et géographes en vallée du Rhône. Anthropos, Paris.

Ebert, J.I., 2000. ‘The State of the Art in “Inductive” Predictive Modeling: Seven Big Mistakes (and Lots of Smaller Ones)’, in:

Wescott, K.J. and R.J. Brandon (eds.), Practical Applications of GIS For Archaeologists. A Predictive Modeling Kit.

Taylor & Francis, London, pp. 129-134.

Favory, F., J.-L. Fiches and S. van der Leeuw, 2003. Archéologie et systèmes socio-environnementaux. Etudes multiscalaires sur la vallée du Rhône dans le programme ARCHAEOMEDES. CRA-Monographies. CNRS Editions, Paris.

(13)

209 CHAPTER 10

Gaffney, V. and M. van Leusen, 1995. ‘Postscript - GIS, environmental determinism and archaeology: a parallel text’, in: Lock, G. and Z. Stančič (eds.), Archaeology and Geographical Information Systems: A European Perspective. Taylor &

Francis, London, pp. 367-382.

Gaffney, V., Z. Stančič and H. Watson, 1995. ‘The impact of GIS on archaeology: a personal perspective’, in: G. Lock and Z.

Stančič (eds.), Archaeology and Geographical Information Systems: A European Perspective. Taylor & Francis, London, pp. 211-229.

Kohler, T.A., 1988, ‘Predictive locational modelling: history and current practice’, in: Judge, W.L. and L. Sebastian (eds.), Quantifying the Present and Predicting the Past: Theory, Method and Application of Archaeological Predictive Modeling. US Bureau of Land Management, Denver, pp. 19-59.

Kuna, M., 1998. ‘The Memory of Landscapes’, in: Neustupný, E. (ed.), Space in Prehistoric Bohemia. Academy of Science, Prague, pp. 77-83.

Kuna, M., 2000. ‘Surface Artefact Studies in the Czech Republic’, in: Bintliff, J., M. Kuna and N. Venclová (eds.), The future of surface artefact survey in Europe. Sheffield Archaeological Monographs 13. Sheffield Academic Press, Sheffield, pp.

29-44.

Kvamme, K.L., 1997. ‘Ranters Corner: bringing the camps together: GIS and ED’. Archaeological Computing Newsletter 47: 1- 5.

Leeuw, S.E. van der (ed.), 1998. The Archaeomedes Project: Understanding the natural and anthropogenic causes of soil degradation and desertification in the Mediterranean Basin. Research Results. Office for Official Publications of the European Communities, Luxembourg.

Leusen, M. van, 2002. Pattern to Process. Methodological investigations into the formation and interpretation of spatial patterns in archaeological landscapes. Rijksuniversiteit Groningen, Groningen. PhD thesis.

Leusen, M. van, J. Deeben, D. Hallewas, P. Zoetbrood, H. Kamermans and Ph. Verhagen, 2002. Predictive modelling for archaeological heritage management in the Netherlands. Baseline report for the BBO research program. Interim report. Rijksuniversiteit Groningen, Groningen.

Leusen, M. van, and H. Kamermans (eds.), in press. Predictive Modelling for Archaeological Heritage Management: A Research Agenda. Nederlandse Archeologische Rapporten. Rijksdienst voor het Oudheidkundig Bodemonderzoek, Amersfoort.

Lock, G. (ed.), 2000. Beyond the Map. Archaeology and Spatial Technologies. NATO Science Series, Series A: Life Sciences - Vol. 321. IOS Press, Amsterdam

Renfrew, C., and E.V. Level, 1979. ‘Exploring dominance: predicting polities from centers’, in: Renfrew, C. and K.L. Cooke (eds.), Transformations: Mathematical approaches to culture change. Academic Press, New York, pp. 145-167.

Ridges, M., in press. ‘Understanding H-G behavioural variability using models of material culture: An example from Australia’, in: Mehrer, M. and K. Wescott (eds.), GIS and Archaeological Site Location Modeling. CRC Press, Boca Raton, Florida USA.

Stančič, Z. and K.L. Kvamme, 1999. ‘Settlement Pattern Modelling through Boolean Overlays of Social and Environmental Variables’, in: Barceló, J.A., I.Briz and A. Vila (eds.), New Techniques for Old Times -CAA98. Computer Applications and Quantitative Methods in Archaeology. BAR International Series 757. Archaeopress, Oxford, pp. 231-237.

Verhagen, P., in press. 'Prospection strategies and archaeological predictive modelling', in: Leusen, M. van and H. Kamermans (eds.), Predictive Modelling for Archaeological Heritage Management: A Research Agenda. NAR, ROB, Amersfoort.

Wheatley, D. 1993, 'Going over old ground: GIS, archaeological theory and the act of perception', in: Andresen, J., T. Madsen, and I. Scollar (eds.), Computing the Past: Computer Applications and Quantitative Methods in Archaeology - CAA 92.

Aarhus. 133-38.

Wheatley, D., 1995. ‘Cumulative viewshed analysis: a GIS-based method for investigating intervisibility, and its archaeological applications’, in: Lock, G. and Z. Stančič (eds.), Archaeology and Geographical Information Systems: A European Perspective. Taylor & Francis, London, pp. 171-185.

Wheatley, D., 1996a. ‘Between the lines: the role of GIS-based predictive modelling in the interpretation of extensive survey data’, in: Kamermans, H. and K. Fennema (eds.), Interfacing the Past. Computer applications and quantitative methods in Archaeology CAA95. Analecta Praehistorica Leidensia 28: 275-292.

Wheatley, D., 1996b. ‘The use of GIS to understand regional variation in Neolithic Wessex’, in: Maschner, H.D.G. (ed.), New methods, old problems: Geographic Information Systems in modern archaeological research. Occasional Paper No. 23.

Center for Archaeological Investigations, Southern Illinois University, Carbondale (IL), pp. 75-103.

Wheatley, D. 2003, 'Making Space for an Archaeology of Place'. Internet Archaeology 15. http://intarch.ac.uk/

journal/issue15/wheatley_index.html

Whitley, T.G., 2000. Dynamical Systems Modeling in Archaeology: A GIS Approach to Site Selection Processes in the Greater Yellowstone Region. Unpublished PhD thesis. Department of Anthropology, University of Pittsburgh, Pittsburgh (PA).

(14)

210 CHAPTER 10

Whitley, T.G., in press. ‘A Brief Outline of Causality-Based Cognitive Archaeological Probabilistic Modeling’, in: Leusen, M.

van and H. Kamermans (eds.), Predictive Modelling for Archaeological Heritage Management: A Research Agenda.

Nederlandse Archeologische Rapporten. Rijksdienst voor het Oudheidkundig Bodemonderzoek, Amersfoort.

Woodman, P.E. and M. Woodward, 2002. ‘The use and abuse of statistical methods in archaeological site location modelling’, in: Wheatley, D., G. Earl and S. Poppy (eds.), Contemporary Themes in Archaeological Computing. Oxbow Books, Oxford.

POSTSCRIPT TO CHAPTER 10

Part of this paper was originally written as a grant proposal for the second phase of the BBO programme. Unfortunately, the research suggested in the paper was not funded, and we have made no further attempts to find other sources of funding. The type of research advocated in this paper is not a priority in Dutch archaeology, and perhaps not even in international archaeology. It is difficult to say why, as the reviews of the grant proposal by external experts were positive, and its scientific and societal relevance was considered high by the review committee. The main objection brought forward against the proposal was the fact that the proposed research could not guarantee a successful outcome, and underestimated the complexity of the matter, so perhaps even needed more funding than was asked for.

However, there is a strong case for doing this type of research, as is explained in the second section of the paper. The post-processual critique of archaeological predictive modelling is mainly based on the conviction that ecological factors cannot offer a full explanation, and therefore not a valid prediction, of site location preferences. This ignores the fact that environmentally based predictive modelling, and related

‘environmental’ methods like site catchment analysis, have been quite successful, provided they use data sets of sufficient quality. But obviously, any predictive model will have a ‘residual’ of sites that do not fit the (environmental) explanatory framework applied, and these are the sites that should be analysed for other factors, including socio-cultural ones. Post-processual theorists however have largely remained silent when it comes to finding a way of integrating socio-cultural factors into predictive modelling. While we are certainly dealing with a complex matter, it seems that earlier attempts to deal with it have focused too much on matters that are truly intangible, like the perception of the landscape in the minds of prehistoric people. Our approach therefore was a more pragmatic one: given the available ‘cultural variables’, can we try to develop predictive models that perhaps will not cover all aspects of site location theory, but that will at least contribute to a better prediction? Unfortunately, we will have no opportunity to find out, at least not in the near future.